The War That No One Waged

August 11, 2025

On August 11, 2025, within a window of one hundred and twenty-two consecutive blocks on the Monero network — from block 3,475,729 to block 3,475,850 — a pool that brought together computing power under the Qubic banner mined sixty-three blocks. That is fifty-one and six tenths percent. Fifty-one percent is the technical threshold past which a pool, while continuing to hold its share, can rewrite the network’s transaction history, reverse confirmed transfers, and censor selected addresses. None of the participants in Qubic did any of that during those twenty-four hours. A few days later, the pool’s share of Monero began to decline — not because it had been pushed out, but because the team itself had shifted priorities to other tasks. No Monero forks happened. No user losses were recorded. On the Kraken exchange, XMR deposits were temporarily suspended for several days, then resumed. The price of Monero pulled back about sixteen percent over the week and started recovering.

If you read only the dry numbers, nothing remarkable happened. One pool briefly exceeded a key technical threshold on one of the networks, and didn’t take advantage of it. But the numbers next to those numbers are worth saying out loud.

Monero’s market cap in August 2025 was about six billion dollars. Qubic’s market cap was about three hundred million. A ratio of one to twenty. A network with twenty times less market cap restructured, over the course of three months — from May to August — a network with twenty times more. Not by attack, not by buying out infrastructure, not by lobbying. By an economic offer. Monero miners moved over to the Qubic pool because it paid, on average, three times more than direct mining on any other pool. They were not deceived. They were not coerced. They saw a better payoff matrix and made a rational decision.

This is the chapter about what exactly happened between May and August 2025, and why what happened matters more than the dry numbers suggest. Because everything this book has said earlier about iteration, memory, transparency, the four properties of Tit-for-Tat, and Axelrod’s three conditions — all of it, before August 2025, was either 1980s mathematics, or simulation modeling, or architectural description. After August 2025, it is an empirical fact on a network with six billion dollars of market cap. That’s where I’ll start.

How the pool grew from two percent to fifty-one

In May 2025, Qubic’s share of Monero’s total hashrate was less than two percent. It was an experimental pool, set up by the Qubic team as a test of one idea — that the computing work the Qubic network produces anyway as part of its own UPoW could simultaneously be used for mining on external networks, in order to convert the rewards back into the Qubic economy. By the end of June, the share had grown to about ten percent. By mid-July — to twenty. By the end of July, Qubic was the number-one Monero pool and reached around forty percent of the hashrate at peak moments. On August 11, fifty-one and six tenths percent in the window of 122 blocks. From two to fifty-one in three months. This needs to be marked as a pace — a pace of migration, not a pace of building infrastructure. The infrastructure was ready in advance: the pool could mine Monero by May. What changed from May to August was not the pool’s code, but the number of miners who chose precisely it. Each of them made an individual decision, and the sum of those decisions added up to a growth curve with a very characteristic shape: a slow start, gradual acceleration, a tipping point, after which new miners come because they hear from other miners about a pool that pays more.

The key moment that explains the pace is the vote that took place inside the Qubic community in early summer 2025. Originally the scheme looked like this: the pool mines Monero, the XMR earned is sold for USDT, the USDT goes to buyback and burning of QU tokens. All the upside from external mining is absorbed by the deflationary mechanism of the Qubic network; a miner who switches into the pool from any other Monero pool earns roughly the same as before, plus an indirect upside through the appreciation of QU that he holds. That was enough to grow the pool from two percent to ten over May–June, but not enough to push past twenty.

After the vote, the scheme changed. Half of the USDT still went to buyback and burning — the deflationary mechanism preserved, but in a softened form. The other half went directly to Qubic’s computor nodes as additional income. That second half is what turned the pool from an experiment into a magnet. After the restructuring, a miner who switched to the Qubic pool started earning, on average, three times more than from direct Monero mining on any other pool. This isn’t a marketing line; it’s a measurable parameter, verifiable from the payouts to specific addresses. The pool honestly paid what it promised, because the promise was built into the code.

Dan Dadybayo, a researcher at Unstoppable Wallet, in an analysis of the situation published in August 2025, formulated it this way: Ivancheglo created a game of incentives in which Monero miners may voluntarily surrender the network if they see a better deal. “Voluntarily surrender the network.” Six words that describe the nature of the event more precisely than most of the technical write-ups from that summer. This is a formulation worth unpacking, because it is not a metaphor. A Monero miner who switched to the Qubic pool in July 2025 did not violate any rules. He did not attack Monero. He did not help attack Monero. He made a market decision — to switch his hashrate to a pool that pays more. The same kind of decision that each of them had already made dozens of times before, when choosing between different Monero pools. From the standpoint of Monero’s rules, the actions of the miners who moved to Qubic are indistinguishable from ordinary migration between pools. From the standpoint of Monero as a coordination system, those actions resulted in one pool accumulating a majority of the hashrate. Not because it forced anyone. Because it offered more. In the first chapter, in its earliest version, I wrote that the new technology “works in parallel, not instead of” — about gas lamps that appeared alongside coal lighting, not in place of it. Back in 2009, that was said about Bitcoin as “Murdoch’s Gas that no one had yet collected into a pipe.” In August 2025, the same principle showed up differently. No one was destroying Monero. Monero as a network kept running, and keeps running; its privacy isn’t broken; its protocol isn’t broken. In parallel with it, a pool appeared in which miners earned three times more, and some of them moved over there. The network didn’t start working worse as a technology because of that; it just temporarily lost the monopoly on its own miners’ attention. After Qubic reduced its activity in Monero, the community migrated to P2Pool, a decentralized pool, and the pluralistic state was restored. The old world works. The new world also works. No one was destroying anyone.

This is parallel architecture in infrastructural form. Not “the better system defeats the worse one through a contest.” But “the better system offers free migration into itself, and some of the participants migrate, and some stay, and that’s normal.” Three conditions at the scale of six billion In the second chapter I went through, in detail, the exact conditions under which the four-line-of-code strategy — Tit-for-Tat — becomes optimal. Axelrod and Hamilton, in their 1981–1984 papers, formulated this as three necessary conditions. The game must be repeated. Participants must remember exactly what their opponent did in the past. Behavior must be visible — either directly, or through a reputational channel. Remove any one of the three, and the optimal strategy shifts from cooperation to pure egoism.

In August 2025, all three conditions were met at industrial scale.

Iteration. Every Monero block is a separate round of the game. A block comes out roughly once every two minutes. Over the three months from May to August, the network ran somewhere on the order of sixty-five thousand rounds. No miner deciding which pool to choose was acting in a one-shot situation. Every choice of a pool was a choice of a partner for the next tens of thousands of rounds, and all participants knew this. The shadow of the future — not as a metaphor, but as a quantitative parameter — was so long that any consideration of momentary gain from cheating became negligible compared to the long-term profit from honest participation.

In the seventh chapter I introduced the concept of an epoch as a round of an iterated game. The discussion there was about Qubic’s own network, where an epoch lasts one week and the composition of computor nodes is updated by the results. Here, in Monero, the same principle works at an even larger scale and on someone else’s network. Every two minutes — a round. Every miner — a player in that round. Every choice of pool — a move in the game, observable by everyone else.

Memory. The Monero blockchain is a public history in which every block is forever recorded together with the address that received the reward for that block. It is impossible to “forget” how many blocks each pool mined last month. It is impossible to misrepresent the situation after the fact: if the Qubic pool had actually paid less than it promised, this would have been immediately visible from the ratio of mined blocks to public payouts. If there had been some hidden channel for redistributing rewards, this violation of the proof-of-work logic would have shown up in orphan blocks or in publication delays. The network’s memory does not depend on the conscientiousness of its participants; it is built into the protocol.

This is a level of memory unattainable in any of the environments Axelrod had in mind when he wrote about real-world analogs of his laboratory conditions — kin groups in primates, small communities, trade guilds. All these systems had memory, but biological and social memory: limited by the number of people capable of personally remembering one another, and by the time those people are alive. The blockchain removes both limitations (this matters!). The network’s memory does not fade with time and does not depend on the number of participants. It is a memory tool that appeared after Axelrod had finished his main work, and appeared not as an implementation of his ideas, but as infrastructure for monetary transfers. But it was precisely this infrastructure that turned out to be the environment in which the conditions of his theory are met strictly.

Transparency. The hashrate of every pool in Monero is publicly visible. You don’t need to be a member of the community, you don’t need a subscription to a specialized service, you don’t need to wait for reports. Anyone with a web browser, anywhere in the world, can go to any of the public Monero monitoring dashboards and see in real time which pool has what share. When the Qubic pool crossed twenty percent in July, this was known within hours. When it crossed forty percent in early August, this was also known within hours. Transparency is not a function of the Monero community, not the result of journalistic work, not the consequence of conscientious disclosures by the pools. It is simply a structural property of a public proof-of-work network. When all three conditions are met, Tit-for-Tat becomes not just an optimal strategy — it becomes what is described in one of Axelrod’s formulations as an “evolutionarily stable strategy.” A strategy the system converges to, given enough rounds. The Monero community reacted instinctively to Qubic’s dominance precisely along the logic of a mirror: miners began moving to P2Pool, discussions of a possible hard fork to change the RandomX algorithm gained traction, the Kraken exchange temporarily suspended XMR deposits. None of these responses were coordinated “political” decisions. They are the spontaneous reaction of a distributed environment in which the three conditions are met, to a temporary disruption of balance. After Qubic reduced activity, the reaction calmly subsided. There were no consequences for miners who continued to work with the pool; no one “punished” them. Monero as a coordination system simply returned to a pluralistic state.

This is industrial validation of the theoretical framework laid down in 1980. Not a controlled experiment in a lab at the University of Michigan. Not a computer simulation. A six-billion-dollar network with thousands of independent participants, in which all three conditions are met strictly, and in which the reaction to disturbance played out exactly as a theory written forty-five years before the event would have predicted. What Qubic could have done and didn’t By early August 2025, when Qubic’s hashrate in Monero had come right up to fifty percent, the project’s team faced an unpleasant technical question. Not “will we hold fifty-one percent” — that was only a question of a few days of work and accumulating hashrate. But “what will we do afterwards, if we actually hold it.” On August 3, 2025, Come-from-Beyond posted, in one of the public channels, a remark that captures the choice the team was facing better than any other material from that summer: “Right. If we wanted we would be DDoSing back. Or even do a PR attack by warning exchanges that once we assert domination we’ll be doing random chain reorganizations while censoring transactions, so any XMR deposit to exchanges would be double-spent with 50% chance. Even DEXes wouldn’t help coz they would be victims of doublespending too. And XMR price would plummet to 0. But, as @dkat said, we frens and won’t do that all.”

This remark is worth unpacking, because it contains six concrete attack vectors, each of which was technically available to the Qubic team in the first week of August 2025, and each of which the team consciously did not use.

The first vector — retaliatory DDoS. From August 3 to August 10, the Qubic team was under a distributed denial-of-service attack that pushed its own network’s hashrate down from two and a half gigahashes per second to eight hundred megahashes at the worst point of the attack. CfB publicly accused one of the key XMRig developers of coordinating the attack; that developer denied it; the source of the attack remained formally unconfirmed. What is confirmed is that the attack went on for more than a week and substantially complicated the holding of hashrate at the needed scale. The Qubic team could have responded symmetrically: use its own infrastructure to coordinate a DDoS attack on its opponents. It didn’t.

Second — a PR attack on exchanges. After crossing the fifty-one-percent threshold, the team could have sent a warning to all exchanges that accepted XMR deposits, saying that from now on any confirmation of a transfer at the exchange should be considered unreliable with a fifty percent probability. This is a technical truth: while holding majority hashrate, a pool can produce chain reorganizations that roll back previously confirmed transactions. Exchanges, on receiving such a warning, would have been forced to defend themselves by temporarily suspending operations with XMR. This would have destroyed Monero’s liquidity on exchanges without a single actually reorganized block — a single public warning from a majority-holding pool would have been enough. Not done.

Third — actual chain reorganizations. Technically the most direct use of the majority position. The pool keeps its blocks secret, accumulating an internal chain longer than the public one, and periodically publishes it, forcing the network to roll back to a new version of history. This method, known as selfish mining, was described in the 2014 paper by Eyal and Sirer; it actually was used as part of the strategy of the final push to fifty-one percent. But selfish mining in the final-capture phase is a tool for accumulating more of one’s own blocks relative to the total. Chaotic reorganizations after the capture are something else: the network turns into an environment in which transaction history stops being stable. A few public reorganizations would have been enough to seriously undermine Monero’s functionality as a payment system. Not done. Fourth — transaction censorship. While holding majority hashrate, a pool can choose which transactions to include in its blocks and which to ignore. A few weeks of such selective mining policy is enough — and any address or class of addresses can be structurally cut off from the ability to confirm transfers. Technically this is not an impossible action; it is regularly discussed as a risk scenario for proof-of-work networks. Not done.

Fifth — an attack on exchange deposits with a fifty-percent probability of failure. This is a combination of the reorganization strategy with exchange mechanics. After holding majority, any XMR deposit sent to any exchange could, with high probability, be reversed through chain reorganization either before or after the exchange credited it to the account. The exchange loses either funds or user trust. On a network with millions of users, such behavior is destructive, and DEX systems do not provide protection, because they too operate on top of the same chain history. Not done.

Sixth — a consequence of the first five, explicitly stated by CfB: the price of XMR will collapse to zero. This is a forecast, not a desire. CfB is not asserting that the price will necessarily fall to zero — he is describing the limit scenario that would result from consistent application of the arsenal. The specific depth of the collapse depends on many variables; what matters in principle is that what was being described was a space of possible outcomes, not a definite result. Not done.

Six attack vectors. Each technically available. Each would have used the pool’s accumulated advantage against its former colleagues on the network. Each remained unused. CfB’s closing line in the same remark, explaining the choice: “as @dkat said, we frens and won’t do that all.” Dkat, the developer of QEarn and lead developer of the Monero operation, had by that point already articulated the team’s position: we don’t finish off partners from whom we’ve gained an advantage.

That Qubic did not destroy Monero is not a consequence of the destruction being impossible. Not a consequence of the team lacking the necessary infrastructure. Not a consequence of external restraint from exchanges, regulators, or the community. It is a decision by the Qubic team, made at the moment of maximum advantage and explicitly recorded in the public record a week before the advantage was realized to its logical limits. Six tools for destroying the network were lying next to the team while it held majority hashrate. The team did not pick up a single one.

Four properties, applied in reality

In the second chapter I went through the four properties of Tit-for-Tat that made it the winner of the 1980 tournaments. The properties were articulated by Axelrod from his analysis of every game played in the first tournament: niceness, retaliation, forgiveness, transparency. Each of them was implemented by Tit-for-Tat through Anatol Rapoport’s four lines of code. Each is theoretically independent of the others, and each is necessary for victory in an environment with the three conditions met. In August 2025, all four properties showed up at industrial scale — not as metaphors, but as observable actions of the Qubic team toward the Monero network and its community. It’s worth fixing this explicitly, because the correspondence is exact. Niceness. Tit-for-Tat never defects first; on the first move, it always cooperates. Qubic did not attack Monero first. Qubic’s original strategy toward Monero was not an attack but an offer: a pool that paid miners more for the same work. On the first move — cooperation in the form of a better economic offer. There was no plan to attack anyone. That, as a result of this offer, a de facto majority of the hashrate ended up captured is a consequence of the offer being convincing to the miners themselves, not of a prior intent to harm the network.

Retaliation. Tit-for-Tat responds to defection immediately — not with delay, not with a margin, not with probability, but exactly on the next round, exactly once. Qubic responded to the DDoS attack that ran from August 3 to August 10. But it did not respond symmetrically — not with a counterattack, not by destroying someone else’s infrastructure, not by punishing specific opponents. It responded by continuing the operation. The hashrate that had dropped to eight hundred megahashes under the attack recovered; the pool held its share in Monero; on August 11, the fifty-one-percent threshold was crossed. That is retaliation in its precise sense — refusal to be punished for the original action. But without escalation and without transferring the punishment onto participants who had nothing to do with the attack.

Forgiveness. Tit-for-Tat doesn’t hold grudges; if the opponent has returned to cooperation, it returns too. Right away. After demonstrating the possibility of capture, the Qubic team did not move to lock in the advantage. The pool’s hashrate in Monero began to decline shortly after August 11 — not because of external pressure, but because of an internal decision to switch compute priorities to DOGE-mining and Aigarth. By August 17 the pool no longer held majority. This action is functionally equivalent to “returning to cooperation”: the team retreated to the network’s pluralistic state, with no external reason to do so other than the fact that the goal of the experiment — to demonstrate the mechanics — had been achieved, and continuing to apply pressure further would have been a transition from testing to actual harm.

Transparency. Tit-for-Tat is predictable: it’s easy for the opponent to understand what it does and why. The Qubic team ran the entire operation in the public field. The plan was announced — that’s visible from posts going back to late May 2025. Technical details were discussed openly, including the August 3 remark about the available attack arsenal. The pool’s hashrate metrics were publicly visible, like those of any other participant in the Monero network. The final report on August 12 on Qubic’s official blog set down the three theses the team considered proven by the experiment, and explicitly announced that the experiment was complete. No hidden channels of coordination; no private negotiations with exchanges; no deals with individual players. Transparency is not a result of honest intentions; it is a structural property of how the operation was run.

Rapoport’s four properties, applied to a real economy at the scale of six billion dollars. Not in a lab, not in a simulation, not in a training environment. On a live network, in real time, with real consequences.

In the second chapter I closed the discussion of Axelrod with the observation that the same formula — “treat others as you would have them treat you” — had been independently arrived at by five people over the course of two and a half millennia. Confucius around 500 BCE. Hillel the Elder in Jerusalem around 30 BCE. Jesus of Nazareth around 30 CE. Kant proved it logically in 1785. Rapoport proved it mathematically in 1980. Five independent discoveries of one formula, with intervals of centuries. On August 12, 2025, this formula received its sixth confirmation — not as a formulation, but as an industrial experiment in which precisely the properties Rapoport sewed into four lines of code showed up exactly as the theory predicted. Bravo, Cfb!

DDoS as a test of the Retaliatory property

It is worth pausing briefly on the August 3–10 episode, because it is the cleanest test of one of the four properties in a live environment.

The DDoS attack against Qubic’s infrastructure began roughly when the pool’s hashrate in Monero reached forty-five percent. Over seven days of attack, the total hashrate of the Qubic network fell from two and a half gigahashes per second to eight hundred megahashes at the worst point. That means Qubic’s own network — the network through which computing work was supposed to be distributed between Moneromining, DOGE-mining, and UPoW tasks for its own consensus — operated for several days in a regime markedly inferior to its usual one. Some peripheral services of the project were temporarily shut down. The financial load on the infrastructure increased. Coordination between computor nodes became more complicated.

And yet the operation in Monero was not interrupted. The Monero hashrate declined, but not critically; the pool held its number-one position for all seven days of the attack. On August 11, when the attack was already winding down, the fifty-one-percent threshold was crossed. The very fifty-one and six tenths percent in the window of 122 blocks that frame the rest of the chapter.

This is a test of the Retaliatory property in its precise formulation. Not “got revenge with the same thing on the same actor” — Qubic didn’t do that. But “kept acting in such a way that the aggression did not achieve its goal.” In the terms of the 1980 tournament: Tit-for-Tat in an environment of aggressive strategies does not surrender and does not retreat; it keeps playing its own game, responding to each act of aggression exactly once — by continuing the cooperative game with those who are ready to cooperate. The Qubic team in August 2025 passed this test literally: it responded to aggression against its own infrastructure by carrying the operation in Monero through to its end, without breaking the chain of public commitments to the miners who had moved into its pool. The twelfth chapter introduced the concept of emergent safety — protection that arose not as an engineered function but as a byproduct of correctly built mechanics. The DDoS episode of August 2025 shows the same phenomenon in reverse: emergent resilience. The Qubic network was not specifically designed to survive a week-long DDoS attack. But an architecture of six hundred and seventy-six nodes distributed across different operators and jurisdictions makes an attack on the infrastructure as a whole exceptionally expensive and not yielding any definite result. Like emergent safety, emergent resilience is a quality that shows up at the moment of crisis, not declared in advance.

What matters in this episode for the broader frame: the network’s ability to withstand coordinated external pressure for a week without interrupting the operation it is currently being used for is no longer a design promise but an observed property. And since the deploy of Aigarth in 2027 will require approximately the same — sustained operation of the infrastructure under load — the data of August 2025 has significance beyond the topic of this chapter.

Incentives Dictate Consensus

On August 12, 2025, Qubic’s official blog published the final report on the Monero experiment. The report set down three theses the team considered empirically proven by the operation:

“Outsourced Computations are Viable: Qubic’s UPoW model is a proven, real-world technology capable of repurposing compute power for external tasks.” “Incentives Dictate Consensus: Any Proof of Work blockchain can be attacked and controlled by providing superior economic incentives to its miners.” “Infrastructure Resilience: Qubic’s decentralized network demonstrated a remarkable ability to withstand and recover from sophisticated, sustained network attacks.”

Three theses. Each of them structural, not evaluative. The first says that the UPoW architecture analyzed in the eighth chapter is capable of actually directing the network’s own computing work to external tasks without losing its main consensus function. The second — that any proof-of-work network can be reoriented by giving its miners better economics, and that this is not an anomaly and not a vulnerability of individual networks, but a structural property of the entire class. The third — that a decentralized network built as a mesh with no single point of failure withstands sophisticated network attacks.

Of the three theses, the second matters most within the frame of this book. “Incentives Dictate Consensus” — Axelrod formulated in the language of the protocol. Any proof-of-work network is a coordination system with a specific payoff matrix. Miners in it make decisions that determine which chain of blocks will be the common consensus. If someone offers miners a better payoff matrix — better specifically along the parameters they are already optimizing for, which is the income from their work — rational miners migrate. This is not betrayal. This is not a moral choice. This is simply a property of an environment in which all three of Axelrod’s conditions — iteration, memory, transparency — are met. In such an environment, cooperation forms around the point where it is economically most profitable. If that point moves — cooperation moves with it.

This means that the question “why does network A, and not network B, withstand a 51% attack” is not a question about the resilience of the consensus algorithm. It is a question about the resilience of the economy. A network survives as long as it is more profitable for its miners to work in it than anywhere else. Once that stops being true, it becomes vulnerable to migration. And this migration cannot be prevented by encryption, by an algorithm, or by a social campaign. It is prevented only by holding the economic superiority of one’s own network.

That is “incentives dictate consensus” in its full form. A structural description of how distributed coordination systems based on the expenditure of computing work operate. On April 13, 2027, by plan, the Aigarth framework moves into production, and the Qubic infrastructure is to provide compute capacity for AI inference as a new type of useful work. The main risk, as CfB put it back on November 12, 2025: “Aigarth depends on miners.” Exactly the same question as in May 2025, but at a different scale. Will miners choose Qubic-network compute tasks — by then, Aigarth tasks rather than Monero-mining — because they pay more? If yes — the deploy works in the regime described as scenario A in the fourteenth chapter. If no — the deploy works technically, but Aigarth applications turn out weak from a lack of compute, and that will be scenario B.

The August 2025 experiment does not answer this question directly. It does not guarantee that April 2027 will see a repeat. But it answers another, more fundamental question: whether the mechanic itself works. Whether a network offering better economics is able to reorient miners without coercion, at the scale of live infrastructure, against competitors’ resistance and under external pressure. The answer to this question has been given. Affirmative. Once, on one specific network, at a known scale. That is not enough to speak of a law. It is enough to speak of a proof of principle.